Content
71%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, codebase-specific skill with excellent concrete templates and strong safety rules (no guessed schemas, listingCapped, contentHash invariants). Its weaknesses are repetition of the meta/runtime split, duplicate reference-implementation sections, and a monolithic single-file structure where deep-dive material could be split into reference files.
Suggestions
Explain the meta.ts / runtime split once (in Directory Structure) and reference it elsewhere; it is currently restated in 'Connector Structure', the meta.ts template commentary, and the Registering section.
Move the selector canonical-pair rules, selector-key table, and content-hash strategy catalog into one-level-deep reference files (e.g. references/selectors.md, references/content-deferral.md), keeping SKILL.md as a concise overview with well-signaled links.
Merge the two 'Reference Implementations' sections (under Content Deferral and near the end) into a single section to remove duplication.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with genuinely non-obvious project rules (canonical pairs, listingCapped reconciliation, contentHash invariants) and avoids generic-concept padding, but it repeats the meta.ts/runtime split across at least four sections ("This mirrors the XBlockMeta... split", "The declarative metadata lives in meta.ts", "Keep meta.ts free of any server/runtime import", and the Registering section) and includes two separate "Reference Implementations" sections. | 3 / 5 |
Actionability | Fully copy-paste-ready scaffolds with exact imports, file paths, registry snippets, helper function names (joinTagArray, parseTagDate, fetchWithRetry, VALIDATE_RETRY_OPTIONS), a selector-key table, and concrete contentHash formats ("gdrive:${file.id}:${file.modifiedTime}"). Placeholder API-call bodies are explicitly justified by the no-guessed-schemas hard rule. | 5 / 5 |
Workflow Clarity | A clear 4-step task sequence, logically ordered sections, an end checklist, and validation checkpoints (validateConfig, validate-selector skill, hard-rule escalation with four fallback options) cover the risky deletion-reconciliation behavior. It falls short of a 5 because the end-to-end build order must be synthesized across sections rather than following one explicit procedure with fix-and-retry feedback loops. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are all absent), so all ~640 lines live in SKILL.md. Sections are well-headed and navigable, but content that belongs in separate reference files — the ~120-line selector canonical-pair deep dive, the content-hash strategy catalog, and the selector-key table — is inlined. | 3 / 5 |
Total | 15 / 20 Passed |